|
XELOPES Library - Technology
Architecture
The XELOPES library conforms completely with the OMG Model Driven Architecture (MDA) standard. The XELOPES core was defined via UML as a CWM expansion and comprehensively  documented. This core forms the platform independent model (PIM) in accordance with MDA specifications.
Various platform-specific models (PSM) were derived and implemented using mappings. There are currently PSMs for Java, C++, C# as well as CORBA and web services. The PSMs are also comprehensively documented.
The XELOPES library features a modular system and contains algorithms from different areas of business intelligence, the focus being on data mining. The algorithms are arranged in packages and the packages can be used to put together flexible Bi applications.
Data import
Data sources for data mining accesses are uniformly modelled using the abstract class MiningInputStream. There are ready-to-use access classes for memories, databases and files including special formats like CSV, Excel and logs. The user can use add-ons to the MiningInputStream class to develop his own data access classes, specifically tailored to his own applications.
Analytical functions
The analytical functions of the XELOPES library are divided into three large packages: multidimensional, data mining and reinforcement learning. The multidimensional package contains multidimensional selections, groupings and a complete OLAP engine. It thus represents an extremely lean implementation for database functions and OLAP. The data mining package contains statistics with multidimensional grouping, decision and regression trees, neural networks, support vector machines, sparse grids, cluster methods, shopping basket analysis with taxonomies and sequence analysis. prudsys has the world's fastest data mining method in the areas of non-linear regression, shopping basket and sequence analysis and sequential shopping basket analysis. The reinforcement learning package contains various methods from the areas of dynamic programming and online learning. It also uses models from the data mining package for approximations.
Results and export
XELOPES stores its models in the CWM-class MiningModel, which can be serialised in various ways. In addition, the models can be exported to other data mining standards like PMML.
|
|
